Network and Predictive Analytics

rm(list = ls())
setwd('~/Desktop/Networks & Predictive Analytics/Project/')
library(igraph)

Attaching package: ‘igraph’

The following objects are masked from ‘package:stats’:

    decompose, spectrum

The following object is masked from ‘package:base’:

    union

de_edges = as.data.frame(read.csv('de_edges.csv', header = TRUE))
de_nodes = as.data.frame(read.csv('de_nodes.csv', header = TRUE))

engb_edges = as.data.frame(read.csv('ENGB_edges.csv', header = TRUE))
engb_nodes = as.data.frame(read.csv('ENGB_nodes.csv', header = TRUE))

es_edges = as.data.frame(read.csv('ES_edges.csv', header = TRUE))
es_nodes = as.data.frame(read.csv('ES_nodes.csv', header = TRUE))

fr_edges = as.data.frame(read.csv('FR_edges.csv', header = TRUE))
fr_nodes = as.data.frame(read.csv('FR_nodes.csv', header = TRUE))

ptbr_edges = as.data.frame(read.csv('PTBR_edges.csv', header = TRUE))
ptbr_nodes = as.data.frame(read.csv('PTBR_nodes.csv', header = TRUE))

ru_edges = as.data.frame(read.csv('RU_edges.csv', header = TRUE))
ru_nodes = as.data.frame(read.csv('RU_nodes.csv', header = TRUE))

all_edges = as.data.frame(read.csv('twitch_edges.csv', header = TRUE))
all_nodes = as.data.frame(read.csv('twitch_nodes.csv', header = TRUE))

de_g <- graph_from_data_frame(de_edges, directed = FALSE, vertices = de_nodes)
engb_g <- graph_from_data_frame(engb_edges, directed = FALSE, vertices = engb_nodes)
es_g <- graph_from_data_frame(es_edges, directed = FALSE, vertices = es_nodes)
fr_g <- graph_from_data_frame(fr_edges, directed = FALSE, vertices = fr_nodes)
ptbr_g <- graph_from_data_frame(ptbr_edges, directed = FALSE, vertices = ptbr_nodes)
ru_g <- graph_from_data_frame(ru_edges, directed = FALSE, vertices = ru_nodes)
all_g <- graph_from_data_frame(all_edges, directed = FALSE, vertices = all_nodes)

C_Names <- c("DE", "ENGB", "ES", "FR", "PTBR", "RU", "ALL")
No_Nodes <- c(length(V(de_g)), length(V(engb_g)), length(V(es_g)), length(V(fr_g)), length(V(ptbr_g)), length(V(ru_g)), length(V(all_g)))
No_Edges <- c(length(E(de_g)), length(E(engb_g)), length(E(es_g)), length(E(fr_g)), length(E(ptbr_g)), length(E(ru_g)), length(E(all_g)))
NW_Diameter <- c(diameter(de_g), diameter(engb_g), diameter(es_g), diameter(fr_g), diameter(ptbr_g), diameter(ru_g), 'NA')
NW_Avg_Dist <- c(mean_distance(de_g), mean_distance(engb_g), mean_distance(es_g), mean_distance(fr_g), mean_distance(ptbr_g), mean_distance(ru_g), 'NA')
NW_Density <- c(edge_density(de_g), edge_density(engb_g), edge_density(es_g), edge_density(fr_g), edge_density(ptbr_g), edge_density(ru_g), 'NA')
NW_Avg_Degree <- c(mean(degree(de_g)), mean(degree(engb_g)), mean(degree(es_g)), mean(degree(fr_g)), mean(degree(ptbr_g)), mean(degree(ru_g)), 'NA')
NW_Cohesion <- c(cohesion(de_g), cohesion(engb_g), cohesion(es_g), cohesion(fr_g), cohesion(ptbr_g), cohesion(ru_g), 'NA')
NW_Compactness <- c(mean(closeness(de_g)), mean(closeness(engb_g)), mean(closeness(es_g)), mean(closeness(fr_g)), mean(closeness(ptbr_g)), mean(closeness(ru_g)), 'NA')
f_info <- data.frame(C_Names, No_Nodes, No_Edges, NW_Diameter, NW_Avg_Dist, NW_Density, NW_Avg_Degree, NW_Compactness)
f_info

rm(list = ls())
library(igraph)
library("RColorBrewer")

pal <- brewer.pal(12, 'Set3')
pal2 <- brewer.pal(11, 'Spectral')

ptbr_edges = as.data.frame(read.csv('PTBR_edges.csv', header = TRUE))
ptbr_nodes = as.data.frame(read.csv('PTBR_nodes.csv', header = TRUE))

ptbr_g <- graph_from_data_frame(ptbr_edges, directed = FALSE, vertices = ptbr_nodes)

ptbr_g

IGRAPH a7f9ff1 UN-- 1912 31299 -- 
+ attr: name (v/c), days (v/n), mature (v/l), views (v/n), partner (v/l), twitch_id (v/n)
+ edges from a7f9ff1 (vertex names):
 [1] ptbr_0   --ptbr_92   ptbr_0   --ptbr_428  ptbr_689 --ptbr_1    ptbr_1147--ptbr_1    ptbr_1666--ptbr_1   
 [6] ptbr_179 --ptbr_2    ptbr_2   --ptbr_587  ptbr_474 --ptbr_2    ptbr_287 --ptbr_2    ptbr_126 --ptbr_2   
[11] ptbr_590 --ptbr_2    ptbr_289 --ptbr_2    ptbr_2   --ptbr_530  ptbr_2   --ptbr_291  ptbr_848 --ptbr_2   
[16] ptbr_267 --ptbr_2    ptbr_1177--ptbr_2    ptbr_2   --ptbr_920  ptbr_2   --ptbr_689  ptbr_2   --ptbr_1785
[21] ptbr_2   --ptbr_923  ptbr_417 --ptbr_2    ptbr_2   --ptbr_1619 ptbr_1040--ptbr_2    ptbr_656 --ptbr_2   
[26] ptbr_2   --ptbr_1432 ptbr_486 --ptbr_2    ptbr_786 --ptbr_2    ptbr_1138--ptbr_2    ptbr_1423--ptbr_2   
[31] ptbr_425 --ptbr_2    ptbr_1365--ptbr_2    ptbr_2   --ptbr_19   ptbr_2   --ptbr_1327 ptbr_1025--ptbr_2   
[36] ptbr_132 --ptbr_2    ptbr_2   --ptbr_866  ptbr_2   --ptbr_1694 ptbr_2   --ptbr_1443 ptbr_608 --ptbr_2   
+ ... omitted several edges

plot(ptbr_g, layout=layout_with_kk, vertex.size= 5, vertex.label= NA, vertex.color = 'skyblue', main="Network Before Filtering")

hist(centralization.degree(ptbr_g, mode='all')$res, breaks = 100, xlab = 'Degree', ylab = 'Frequency', main = 'Histogram of Degree Centrality\n Before Filtering')

ptbr_nodes_min <- ptbr_nodes[order(-ptbr_nodes$views),][1:100,]
ptbr_edges_min <- ptbr_edges[which(ptbr_edges$from %in% ptbr_nodes_min$id & ptbr_edges$to %in% ptbr_nodes_min$id),]

par(mfrow=c(1,2), mar=c(1,1,1,1))
ptbr_g_min <- graph_from_data_frame(ptbr_edges_min, directed = FALSE, vertices = ptbr_nodes_min)
plot(ptbr_g, layout=layout_with_kk, vertex.size= 5, vertex.label= NA, vertex.color = 'skyblue', main="Network Before Filtering")
plot(ptbr_g_min, layout=layout_with_kk, vertex.size= 8, vertex.label= NA, vertex.color= 'skyblue', main="Network After Filtering")

as_data_frame(ptbr_g_min, what="vertices")

ptbr_colnames <- c("Size", "No_Edges", "Diameter", "Average Distance", "Degree Centrality", "SD Degree Centrality", "Closeness Centrality", "SD Closeness Centrality", "Betweenness Centrality", "SD Betweenness Centrality", "Density", "Average Degree", "Cohesion", "Compactness", "Clustering Coefficient")

ptbr_values <- c(length(V(ptbr_g_min)), length(E(ptbr_g_min)), diameter(ptbr_g_min), mean_distance(ptbr_g_min), centralization.degree(ptbr_g_min)$centralization, sd(centralization.degree(ptbr_g_min)$res), centralization.closeness(ptbr_g_min)$centralization, sd(centralization.closeness(ptbr_g_min)$res), centralization.betweenness(ptbr_g_min)$centralization, sd(centralization.betweenness(ptbr_g_min)$res), edge_density(ptbr_g_min), mean(degree(ptbr_g_min)), cohesion(ptbr_g_min), mean(closeness(ptbr_g_min)), transitivity(ptbr_g_min,type="global"))
ptbr_info <- data.frame(ptbr_colnames, ptbr_values)
colnames(ptbr_info) <- c('Metric','Values')
ptbr_info

par(mfrow=c(1,2), mar=c(2,2,2,2))
hist(centralization.degree(ptbr_g, mode='all')$res, breaks = 100, xlab = 'Degree', ylab = 'Frequency', main = 'Histogram of Degree Centrality\n Before Filtering')
hist(centralization.degree(ptbr_g_min, mode='all')$res, breaks = 100, xlab = 'Degree', ylab = 'Frequency', main = 'Histogram of Degree Centrality\n After Filtering')

set.seed(1000)
par(mfrow = c(1, 2), mar = c(2, 2, 2, 2))
c_g2 <- fastgreedy.community(ptbr_g_min)
res_g2 <- simplify(contract(ptbr_g_min, membership(c_g2)), remove.multiple = TRUE, remove.loops = TRUE)
plot(res_g2, vertex.size= 10, vertex.label = NA, layout=layout_with_kk, vertex.color=pal, main="FastGreedy Community")

c_g3 <- cluster_louvain(ptbr_g_min)
res_g3 <- simplify(contract(ptbr_g_min, membership(c_g3)), remove.multiple = TRUE, remove.loops = TRUE)
#plot(res_g3, vertex.size= 10, vertex.label = NA, main = length(res_g3), layout=layout_with_kk, vertex.color=pal)

c_g4 <- spinglass.community(ptbr_g_min)
res_g4 <- simplify(contract(ptbr_g_min, membership(c_g4)), remove.multiple = TRUE, remove.loops = TRUE)
plot(res_g4, vertex.size= 10, vertex.label = NA, main = "Spinglass Community", layout=layout_with_kk, vertex.color=pal)

c_g5 <- walktrap.community(ptbr_g_min)
res_g5 <- simplify(contract(ptbr_g_min, membership(c_g5)), remove.multiple = TRUE, remove.loops = TRUE)
#plot(res_g5, vertex.size= 10, vertex.label = NA, main = length(res_g5), layout=layout_with_kk, vertex.color=pal)

c_g_groups2 <- groups(c_g2)
sort(lengths(c_g_groups2), decreasing = TRUE)

 1  3  2 
51 43  6 

c_g_groups4 <- groups(c_g4)
sort(lengths(c_g_groups4), decreasing = TRUE)

 3  1  2 
39 33 28 

par(mfrow = c(1, 2), mar = c(2, 2, 2, 2))
plot(ptbr_g_min, vertex.color=pal[membership(c_g2)], layout=layout_with_kk, vertex.size= 10, vertex.label.cex = 0.5, main="Fast Greedy Community")
plot(ptbr_g_min, vertex.color=pal[membership(c_g4)], layout=layout_with_kk, vertex.size= 10, vertex.label.cex = 0.5, main="SpinGlass Community")

library(threejs)

Registered S3 methods overwritten by 'htmltools':
  method               from         
  print.html           tools:rstudio
  print.shiny.tag      tools:rstudio
  print.shiny.tag.list tools:rstudio
Registered S3 method overwritten by 'htmlwidgets':
  method           from         
  print.htmlwidget tools:rstudio

net <- ptbr_g_min
V(net)$color <- pal[membership(c_g2)]
V(net)$size <- 5
E(net)$color = 'black'

graphjs(net, showLabels = T)

library(threejs)

net <- ptbr_g_min
V(net)$color <- pal[membership(c_g4)]
V(net)$size <- 5
E(net)$color = 'black'

graphjs(net, showLabels = T)

par(mfrow=c(1,2), mar=c(1,1,1,1))
ptbr_g_min_dcent <- delete_vertices(ptbr_g_min, "ptbr_382")

V(ptbr_g_min_dcent)$dcent <- centralization.degree(ptbr_g_min_dcent)$res
V(ptbr_g_min_dcent)$color <- ifelse(V(ptbr_g_min_dcent)$dcent > 49, pal2, 'white')

ptbr_g_min_dcent_tt <- induced_subgraph(ptbr_g_min_dcent, V(ptbr_g_min_dcent)[which(V(ptbr_g_min_dcent)$dcent > 49)])

plot(ptbr_g_min_dcent, vertex.size=V(ptbr_g_min_dcent)$dcent*0.25, vertex.label= NA, layout=layout_with_fr, main="Degree Centrality")

plot(ptbr_g_min_dcent_tt, vertex.label.cex= 1, vertex.size=V(ptbr_g_min_dcent_tt)$dcent*0.30, layout=layout_with_fr, main="Degree Centrality for Top 11")

par(mfrow=c(1,2), mar=c(1,1,1,1))
ptbr_g_min_evcent <- delete_vertices(ptbr_g_min, "ptbr_382")

V(ptbr_g_min_evcent)$evcent <- eigen_centrality(ptbr_g_min_evcent)$vector
V(ptbr_g_min_evcent)$color <- ifelse(V(ptbr_g_min_evcent)$evcent > 0.8589211, pal2, 'white')

ptbr_g_min_evcent_tt <- induced_subgraph(ptbr_g_min_evcent, V(ptbr_g_min_evcent)[which(V(ptbr_g_min_evcent)$evcent > 0.8589211)])

plot(ptbr_g_min_evcent, vertex.size=V(ptbr_g_min_evcent)$evcent*15, vertex.label= NA, edge.arrow.size=.4, layout=layout_with_fr, main="EigenVector Centrality")

plot(ptbr_g_min_evcent_tt, vertex.size=V(ptbr_g_min_evcent_tt)$evcent*20, vertex.label.cex= 1, layout=layout_with_fr, main="EigenVector Centrality for Top 10")

par(mfrow=c(1,2), mar=c(1,1,1,1))
ptbr_g_min_ccent <- delete_vertices(ptbr_g_min, "ptbr_382")

V(ptbr_g_min_ccent)$ccent <- centralization.closeness(ptbr_g_min_ccent)$res
V(ptbr_g_min_ccent)$color <- ifelse(V(ptbr_g_min_ccent)$ccent > 0.6666667, pal2, 'white')

ptbr_g_min_ccent_tt <- induced_subgraph(ptbr_g_min_ccent, V(ptbr_g_min_ccent)[which(V(ptbr_g_min_ccent)$ccent > 0.6666667)])

plot(ptbr_g_min_ccent, vertex.size=V(ptbr_g_min_ccent)$ccent*15, vertex.label= NA, edge.arrow.size=.4, layout=layout_with_fr, main="Closeness Centrality")

plot(ptbr_g_min_ccent_tt, vertex.size=V(ptbr_g_min_ccent_tt)$ccent*30, vertex.label.cex= 1, layout=layout_with_fr, main="Closeness Centrality for Top 11")

ptbr_g_min_bcent <- delete_vertices(ptbr_g_min, "ptbr_382")

g.mat <- as.matrix(get.adjacency(ptbr_g_min_bcent))
g.bc <- sna::betweenness(g.mat)
plot(ptbr_g_min_bcent, vertex.color= "skyblue", vertex.size=g.bc*0.05, vertex.label= NA, layout=layout_with_fr, main="Betweenness Centrality")

library(keyplayer)

Attaching package: ‘keyplayer’

The following object is masked from ‘package:igraph’:

    contract

ptbr_g_min_am <- as.matrix(get.adjacency(ptbr_g_min))
ptbr_g_min_fragment <- kpset(ptbr_g_min_am, size = 10, type = "fragment")
V(ptbr_g_min)$color <- "skyblue"
V(ptbr_g_min)$color[ptbr_g_min_fragment$keyplayers] <- "salmon"
plot(ptbr_g_min, mark.groups = ptbr_g_min_fragment$keyplayers, mark.col = NA, mark.border = "black",
    vertex.size = 10, vertex.label.cex = 0.3, vertex.label.color = "black",
    edge.arrow.size = 0.25, layout = layout_with_kk, main = "Network w/ Key Players")

library(blockmodeling)

To cite package 'blockmodeling' in publications please use package citation and (at least) one of the articles:

  Žiberna, Aleš (2007). Generalized blockmodeling of valued networks. Social Networks 29(1), 105-126.

  Žiberna, Aleš (2008). Direct and indirect approaches to blockmodeling of valued networks in terms of regular
  equivalence. Journal of Mathematical Sociology 32(1), 57–84.

  Žiberna, Aleš (2014). Blockmodeling of multilevel networks. Social Networks 39, 46–61.
  https://doi.org/10.1016/j.socnet.2014.04.002.

  Žiberna, Aleš (2022).  Generalized and Classical Blockmodeling of Valued Networks, R package version 1.1.3.

To see these entries in BibTeX format, use 'print(<citation>, bibtex=TRUE)', 'toBibtex(.)', or set
'options(citation.bibtex.max=999)'.

rege2<-REGE.ownm.for(M=ptbr_g_min_am)$E 
clu <- cutree(hclust(d=as.dist(1-rege2),method="ward.D"), k=3)
V(ptbr_g_min)[names(clu)]$color <- clu
plot(ptbr_g_min, vertex.color=V(ptbr_g_min)$color, vertex.size=10, vertex.label.cex=0.01, main = "Twitch PTBR top 100 Streamers REGE plot")